Tim said:
I'd hesitate to say that this is impossible, but with volts driving
the speaker it'd be very difficult to accurately sense the millivolts
of signal that you wanted to listen to. If it can be done at all you
would need to know some pretty heavy DSP for the adaptive filter,
you'd have to plan on spending man-months, and you wouldn't get good
audio quality when you're done.
So don't think you can throw it together in a weekend, but go for it
if you really want to.
What you mainly get is the back EMF at resonance frequency, where the
efficiency of the speaker is highest.
Imagine the normal 0.1 to 1% efficiency will give you a very small signal.
Compared to the driving signal it will be 60 to 80dB smaller(depending on
the radiating area) plus needs to be filtered with a notch filter because of
the mentioned fundamental resonance. Any irregularity in speaker impedance
will be magnified by the same amount, so the result will be sounding awful,
even if you manage to get something out at all. The sensing resistor in the
speaker feed will eat up on amplifier power and modify low frequency
response.
With a small microphone without the irregularities of the speaker impedance
it will still be difficult to eliminate the direct speaker signal and get a
useful output if it is near the driving speaker. Since the sound pressure
level doubles every half of the distance it is *much* higher near the
transmitter. So I would start with this challenge. And it would be mostly
the room reflections what you get, which still have to be eliminated.
Remember the sound when somebody uses these "hands-free" phones?